In human-robot interaction, sensors are relevant in guaranteeing stability and high performance in real-time applications. Nonetheless, accuracy and portable sensors for robots usually have high costs and little flexibility to process signals with free software. Therefore, we propose a wearable sensor network to measure lower limb angular position in human-robot interaction systems. The methodology employed to achieve the aim consisted in implementing a wireless network using low-cost devices, verifying design requirements, and making a validation via a proof of concept. The requirements to design the network include low loss of information, real-time communication, and sensor fusion to estimate the angular position using a gyroscope and accelerometer. Hence, the sensor network developed has a client-server architecture based on ESP8266 microcontrollers. In addition, this network uses the standard 802.11 b/g/n to transmit angular velocity and acceleration measures. Furthermore, we implement the user datagram protocol (UDP) protocol to operate in real-time with a sample time of 10 ms. Finally, we implement a proof of concept to show the system’s effectiveness. Thus, we use the Kalman filter to estimate the angular position of the foot, shin, thigh, and hip. Results indicate that the implemented sensor network is suitable for real-time robotic applications.

1. TELKOMNIKA Telecommunication Computing Electronics and Control
Vol. 21, No. 2, April 2023, pp. 390~399
ISSN: 1693-6930, DOI: 10.12928/TELKOMNIKA.v21i2.23483  390
Journal homepage: http://telkomnika.uad.ac.id
Wearable sensor network for lower limb angle estimation in
robotics applications
Andres L. Jutinico1,2
, Gustavo A. R. Rodriguez2
, Julian Rolando Camargo Lopez1
1
Department of Electronic Engineering, Faculty of Engineering, Universidad Distrital Francisco José de Caldas, Bogotá, Colombia
2
Master of Information Sciences and Communications, Faculty of Engineering, Universidad Distrital Francisco José de Caldas, Bogotá,
Colombia
Article Info ABSTRACT
Article history:
Received Mar 02, 2022
Revised Sep 27, 2022
Accepted Oct 26, 2022
In human-robot interaction, sensors are relevant in guaranteeing stability and
high performance in real-time applications. Nonetheless, accuracy and
portable sensors for robots usually have high costs and little flexibility to
process signals with free software. Therefore, we propose a wearable sensor
network to measure lower limb angular position in human-robot interaction
systems. The methodology employed to achieve the aim consisted in
implementing a wireless network using low-cost devices, verifying design
requirements, and making a validation via a proof of concept.
The requirements to design the network include low loss of information,
real-time communication, and sensor fusion to estimate the angular position
using a gyroscope and accelerometer. Hence, the sensor network developed
has a client-server architecture based on ESP8266 microcontrollers.
In addition, this network uses the standard 802.11 b/g/n to transmit angular
velocity and acceleration measures. Furthermore, we implement the user
datagram protocol (UDP) protocol to operate in real-time with a sample time
of 10 ms. Finally, we implement a proof of concept to show the system’s
effectiveness. Thus, we use the Kalman filter to estimate the angular position
of the foot, shin, thigh, and hip. Results indicate that the implemented sensor
network is suitable for real-time robotic applications.
Keywords:
Angle estimation
IEEE 802.11 standard
Kalman filter
Wearable systems
Wireless sensor network
This is an open access article under the CC BY-SA license.
Corresponding Author:
Andres L. Jutinico
Department of Electronic Engineering, Faculty of Engineering
Universidad Distrital Francisco José de Caldas, Bogotá, Colombia
Email: aljutinicoa@udistrital.edu.co
1. INTRODUCTION
Stroke is considered a frequent neurological emergency and constitutes the second leading cause of
death globally, with 12.2 million cases per year and 6.55 million losses per year, respectively [1], [2].
In addition, this disease has become the leading cause of long-term disability affecting daily life activities.
Moreover, this medical condition entails a socioeconomic impact with high costs in initial medical care and
enormous difficulties for patients who usually need treatment and rehabilitation.
Research on robotics rehabilitation has put its efforts at the service of health. There are robots
employed in human care, which are used to recover the activity of the locomotive systems of the human
being. Examples of these technologies are exoskeletons and robotic platforms for rehabilitation [3]. In this
sense, robots can support physical therapist work and help the user repeat training with a controlled and
monitored intensity to promote neurogenesis and motor control [4]. Electronic instrumentation is an area to
be applied in this type of robot, which is growing and contributes to diminishing the cost of the same, having
a positive impact on society, and allowing for deeper studies or significant scientific contributions [5].

2. TELKOMNIKA Telecommun Comput El Control 
Wearable sensor network for lower limb angle estimation in robotics applications (Andres L. Jutinico)
391
Commercially there are systems for acquiring kinematic and biomechanical signals used in robotic
rehabilitation. However, these devices have high costs and little flexibility to process signals with free software.
This article proposes a wearable sensor network to measure lower limb angular position in human-robot
interaction systems. The network consists of three inertial measurement units (IMU) to determine limb
acceleration and angular velocity. The angular position is estimated and processed by ESP8266
microcontrollers. Furthermore, this sensor network uses the 802.11b/g/n standard for interconnection with a
host by the user datagram protocol (UDP). Finally, we validate the system operation by a proof of concept,
estimating the angular position through the Kalman filter.
Industrial communications are used to supervise physical, chemical, and environmental variables,
among others. In this kind of system, wired networks guarantee the integrity of the data. Therefore, nodes can
work in a distributed way and using specific protocols, such as standard wired communication protocols:
RS-485, highway addressable remote transducer (HART), Profibus, Ethernet, and Modbus. In robotic
applications, the selection of communication systems is also given by guaranteeing data integrity.
Nevertheless, the device’s portability plays an essential role in adjusting sensor networks with mechatronic
devices. Currently, communication systems use several protocols for transferring information over wireless
means with different bandwidths and frequencies of operation, highlighting communication protocols such as
Bluetooth, Zigbee, and WiFi, among others [6], [7].
Wireless sensor networks include a set of devices to measure variables of a process and transmit
them over small or long distances [8], [9]. Sensor networks are regularly provided with a server or a base
station that serves as a reference point for receiving information, operating in most cases with homogeneous
devices according to the nodal distribution. These nodes generally can process and store information and
have various sensors. However, they do not necessarily have to be the same. Conventionally, nodes have
sensors, controllers, communication modules, and microprocessors. Depending on the purpose of
development, the processing applications will have varying levels of complexity, either to capture the
information or to analyze it [10].
Around the world, there are prototypes and commercial systems that focus on motor rehabilitation
using robotics. These robots use a wide variety of communication protocols for sensors. Biofeedback systems
are widely used in patients with balance disorders due to vestibular loss, stroke, or aging. In addition, patients
often undergo continuous balance training regimens with the help of biofeedback systems that can effectively
increase postural stability. For example, Alahakone and Senanayake [11] employ a wireless inertial sensor
for trunk tilt angle measurements, a complete software with an interface for data acquisition, processing, and
visualization. The system allows video recording, feedback signal generation, data analysis, and portable
vibrating actuators to provide feedback. In [12] developed the inertia-based detection modality with increased
biofeedback to study trunk swing and a training protocol for clinical evaluation. In [13] show a wireless
insole sensor biofeedback and gait analysis system to rehabilitate balance control for post-stroke patients.
The system integrates a wireless network of home and open-source electronic healthcare sensors to be used
with mobile devices. In [14] highlight ZigBee’s higher communication efficiency and lower power
consumption. In addition, they show a comparison between ZigBee and Bluetooth serial port profiles for
wireless body-area sensor networks. The system called Kimera is a portable brain-computer interface (BCI)
with wireless information transfer mechanisms that consists of two layers of hardware architecture, data
acquisition, and wireless transmission by a Bluetooth ARM chip [15]. In [16] present a wearable multi-sensor
system for human motion monitoring for use in rehabilitation. This system integrates five small wireless
modules, interconnected with accelerometer data at high frequency, synchronized by an external master
device, making the system ideal for patient monitoring. Wireless biomedical implants are the subject of
research in wireless body area networks (WBAN). For example, the incremental relaying communication
protocol based on mutual information (MI) is presented [8]. The communication path selection is given by
the deeply implanted and the receiver distance. The results show that the MI-based incremental
retransmission technique achieves better performance compared to other proposed techniques. In [17] show a
study to validate the accuracy of the inertial measurement unit-based knee angle measurement system during
physiotherapy exercises. The used system has a mobile application for recording these data.
Robots for rehabilitation must measure biosignals to promote motor control and safe in human-robot
interaction. In this context, sensors must couple on the robot and the human and transfer data by a network.
There are several examples of this: the robotic training system for upper limb rehabilitation counts with an
exoskeleton that allows the therapist to modify the patient’s routine. The robot supports various trajectories
and repetitions followed by the exoskeleton in the shoulder, elbow, and wrist joints. The robot has sensors for
angular position and electromyographic (EMG) signals and a user interface developed in LabView [18].
NOTTABIKE is a robot designed to investigate and promote lower limb motor control by using a platform
that owns a bike for exercise. This device has a servo motor, angular encoders, pedals, force-torque sensors,
and an EMG system [19]. In [20] shows the design and control of an active knee orthosis driven by a rotary
series elastic actuator (SEA).

3.  ISSN: 1693-6930
TELKOMNIKA Telecommun Comput El Control, Vol. 21, No. 2, April 2023: 390-399
392
The custom rotary SEA is responsible for driving an active knee orthosis and allowing the torque
measurement. Then, the user attaches the robot to its body and executes controlled movements [21]. The system
robotic platform for ankle rehabilitation uses a SEA for control and force measurement, a sensor for the angular
position, and EMG. In [4], [22] introduce a Markovian force and impedance control for human-robot
interaction in this type of robot. The exoskeleton reported in [23] is a mechatronic device to assist in passive
physiotherapy. The robot follows computer-generated trajectories of knee and ankle joints. Peña et al. [20]
propose an EMG-driven adaptive impedance control to improve user engagement during the rehabilitation
session by considering an active knee orthosis. Another example of EMG sensing where Sartori et al. [24]
used EMG data to build a musculoskeletal model to predict stiffness in the human ankle and knee joints.
The contribution of the estimation of angular positions in robotic rehabilitation improves the
performance and efficiency of therapies. Some research has used the Kalman filter for this aim. In [25] is
shown a standard angular estimation approach by Kalman filters for lower limb exoskeletons for inertial
measurement unit (IMU) based on individual link configurations. In [26], introduce a global approach to
estimate absolute angles of the lower limb segments by the Kalman filter. This approach uses inertial sensors
and the exoskeleton encoder facilitating gait analysis and exoskeleton control.
2. METHOD
2.1. Communication topology
Figure 1 shows the overall configuration of the proposed network sensors. The system uses three
nodes, each with the inertial measurement unit (IMU) sensor and a microcontroller with a communication
wifi module. The sensors are located at the foot, shin, femur, and hip height and transmit sensors data to a
server implemented in a microcontroller. The communication system uses the UDP protocol to transfer the
information to a computer that stores the data in files for collection and processing.
UDP protocol is in the transport layer of the open systems interconnection model (OSI model) and can
be used according to standard 802.11. The protocol provides a datagram exchange service over the network;
therefore, it cannot ensure the delivery of datagrams or packets. In this project, the UDP protocol was selected
since it is not possible to carry out measurement retransmissions due to the strict delay requirements of the
system. Each node (is to say, the clients) counts with an inertial measurement unit MPU6050 and an ESP8266
microcontroller, which has an 802.11 communication module. Besides, the server uses an ESP8266
microcontroller to establish wireless communication and send the data to the computer by a USB wire. Notice
that the information sent from sensors in UDP datagrams does not establish a two-way connection to the server
since we are looking for real-time one-way and not the arrival confirmation. Figure 2 shows the workflow to
save data from the sensors. First, devices are started. The following is to establish the connection between nodes
and the server. Then to accomplish this requirement, there is a timeout condition for server response until 30 s.
Once establish this connection each sensor is linked and begins data detection, and data values are saved.
Figure 1. The overall configuration of the proposed network sensors

4. TELKOMNIKA Telecommun Comput El Control 
Wearable sensor network for lower limb angle estimation in robotics applications (Andres L. Jutinico)
393
Figure 2. Workflow to save data from the sensors
2.2. Mathematical model of acceleration and angular velocity
IMU are attached to each segment of the body as shown in Figure 3. The angular velocity in each
limb is modeled as:
𝜃̇(𝑡) = 𝜃𝑔
̇ (𝑡) + 𝑏𝑔(𝑡) + 𝜂𝑔(𝑡) (1)
𝑏̇(𝑡) = −
1
𝜏𝑔
𝑏𝑔(𝑡) + 𝜂𝑏𝑔(𝑡) (2)
Where 𝜃𝑔
̇ (𝑡) is the measurement angular velocity from the gyroscope, 𝜂𝑔(𝑡) is the Gaussian noise with
variance 𝜎𝑔
2
, 𝑏𝑔(𝑡) representing the bias of the gyroscope, 𝜏𝑔 is the Markov process correlation time and 𝜂𝑏𝑔
is the white Gaussian noise with variance 𝜎𝑏𝑔
2
.
Figure 3. Location of IMU
The acceleration of each link, 𝑎(𝑡), in an instant of time, 𝑡, is modeled as:
𝑎(𝑡) = 𝑎𝑎(𝑡) + 𝑏𝑎(𝑡) + 𝜇𝑎(𝑡) (3)
𝑏̇𝑎(t) = −
1
τa
ba(t) + μ𝑏𝑎(t) (4)
Where 𝑎𝑎(𝑡) is the acceleration obtained from the accelerometer on the 𝑥-axis of the local coordinate system,
𝜇𝑎(𝑡) is the Gaussian noise with variance 𝜎𝑔
2
, 𝑏𝑎(𝑡) is the bias of the accelerometer, 𝜏𝑎 is the Markov process
correlation time and 𝜇𝑏𝑎(𝑡) is white Gaussian noise with variance 𝜎𝑏𝑎
2
.
2.3. Strategy for sensor fusion
In this section, the angular position estimation of each link is obtained from sensor fusion, using
measures of acceleration and angular velocity. In this paper, the selected method to make sensor fusion is the
Kalman filter. This filter uses the gyroscope signal as the primary source for position estimation. The
accelerometer signal is considered redundant information to correct the estimated value of the gyroscope that is
affected by integration errors. The angular velocity estimated by the gyroscope and the bias are modeled as:
𝜃
̂̇
𝑔(𝑡) = 𝜃̇𝑔(𝑡) + 𝑏
̂𝑔(𝑡) (5)

5.  ISSN: 1693-6930
TELKOMNIKA Telecommun Comput El Control, Vol. 21, No. 2, April 2023: 390-399
394
𝑏
̂̇
𝑔(𝑡) = −
1
τ𝑔
𝑏
̂𝑔(𝑡) (6)
Therefore, the difference between actual and estimated values is:
Δθ̇(𝑡) = θ̇(𝑡) − θ
̂̇(𝑡)
= 𝑏𝑔(𝑡) − 𝑏
̂𝑔(𝑡) + η𝑏𝑔(𝑡)
= Δ𝑏𝑔(𝑡) + η𝑏𝑔(𝑡) (7)
And,
Δ𝑏̇𝑔(𝑡) = 𝑏̇𝑔(𝑡) − 𝑏
̂̇
𝑔(𝑡)
= −
1
τ𝑔
(𝑏𝑔(𝑡) − 𝑏
̂𝑔(𝑡)) + η𝑏𝑔(𝑡)
= −
1
τ𝑔
Δ𝑏𝑔(𝑡) + η𝑏𝑔(𝑡) (8)
Thus, the state space representation of the estimation system is given by:
[
Δθ̇(𝑡)
Δ𝑏𝑔
̇ (𝑡)
] = [
0 1
0 −
1
τ𝑔
] [
Δθ(𝑡)
Δ𝑏𝑔(𝑡)
] + [
1 0
0 1
] [
η𝑔(𝑡)
η𝑏(𝑡)
] (9)
𝑥̇ = 𝐴𝑥(𝑡) + 𝐵ω(𝑡) (10)
Where 𝛥𝜃(𝑡) and Δ𝑏𝑔(𝑡) are the position estimation and offset errors of the gyroscope, respectively. The
states of the systems are 𝑥, 𝜔 is the white Gaussian noise, and 𝐴 and 𝐵 are parametric matrices that represent
the model (9). The angular position obtained from the accelerometer is given by:
𝜃
̂𝑎(𝑡) = arcsin (
𝑎𝑎
𝑔𝑒
) = θ(𝑡) + η𝑎(𝑡) (11)
Where 𝑔𝑒 is the gravitational acceleration value on the 𝑥 axis. Notice that no is considered dynamic
accelerations. The output equation of the state space representation is given by the difference between the
estimated position of the accelerometer and the estimated position of the gyroscope.
𝑧(𝑡) = 𝜃
̂𝑎(𝑡) − 𝜃
̂𝑔 (𝑡),
= 𝜃(𝑡) + 𝜂𝑎(𝑡) − 𝜃
̂𝑔(𝑡) (12)
= Δθ(𝑡) + η𝑎(𝑡).
Thus, the estimated values for the angular position and gyroscope error are given by:
𝑧 = 𝐶𝑥(𝑡) + 𝑣(𝑡) = [1 0] [
Δθ(𝑡)
Δ𝑏𝑔(𝑡)
] + η𝑎(𝑡) (13)
θ
̂(𝑡) = 𝜃
̂𝑔(𝑡) + Δ𝜃
̂(𝑡) (14)
𝑏
̂(𝑡) = 𝑏
̂𝑔(𝑡) + Δ𝑏
̂𝑔(𝑡) (15)
Where 𝐶 is a parametric matrix, 𝑣 represent a random error, and Δ𝜃
̂(𝑡)and Δ𝑏
̂𝑔(𝑡) are estimated angular
position values and compensation errors, respectively.
3. RESULTS AND DISCUSSION
To validate the developed system, we made a test where a user walked for a straight line of 10 meters.
The test had an approximate duration of 12 seconds. The trial started once the communication system was
linked by wifi to the UDP server. Sensors were located at the foot, shin, femur, and hip height, as is shown in
Figure 4(a) and Figure 4(b) shows the electronic device for each node.

6. TELKOMNIKA Telecommun Comput El Control 
Wearable sensor network for lower limb angle estimation in robotics applications (Andres L. Jutinico)
395
(a) (b)
Figure 4. Test for system validation: (a) ten meter test and (b) node
3.1. Angular position estimation using Kalman filter
To validate the system, we perform a proof of concept. For this proof, a healthy user walked for ten
meters while the sensors network transmitted each limb’s acceleration and angular velocity. Furthermore, we
compared the estimation results from the Kalman filter with the estimation given by a complementary filter.
For the calculation of the angular position from the speed measurement performed by the gyroscope, it is
necessary to consider the resolution of the sensor. Therefore, the following conversion factor is used:
𝐺𝑅 =
2𝑛
250𝑜/𝑠
(16)
Where 𝑛 = 16 is the number of bits used by the sensor, and 250 is a parameter taken from the sensor
datasheet. Therefore, 𝜔(𝑘) = 𝜔
̅(𝑘)/𝐺𝑅, where ω
̅ is the gyroscope measure. The angular position can be
performed from the following integrative recursion.
θ𝑥𝑔(𝑘) = τ𝑠ω(𝑘 − 1) + θ𝑥𝑔(𝑘 − 1) (17)
We use a sample time 𝑇𝑠 = 10 ms. Figure 5 shows the 𝜃𝑥𝑔(𝑘) angle estimation for the foot sensor
during the experiment. We highlight that the velocity integration of the gyroscope causes an angular bias.
For the calculation of the angular position from the measurement of the acceleration performed by the sensor,
the resolution of the sensor should be considered, therefore the following conversion factor is used:
𝐴𝑅 =
2𝑛
4×9.81𝑚/𝑠2 (18)
Where 𝑛 = 16 is the number of bits used by the IMU, and 𝑔 = 9.81 m/s2
is the gravity. Thus,
𝑎𝑥(𝑘) = 𝑎
̅𝑥(𝑘)/𝐴𝑅, where 𝑎
̅𝑥 is the acelerometer measure. The angular position relative to the 𝑥-axis can be
estimated according with (11), as:
𝜃𝑥𝑎 = 𝑎𝑟𝑐𝑠𝑖𝑛(𝑎𝑥/𝑔) (19)
Figure 5 shows the result of estimating the angle from accelerometer signal 𝜃𝑥𝑎. From the bottom
figure, it is appreciated that, although the signal has no bias, it is characterized by a large amount of noise.
This graph shows the need for sensory fusion to perform a more accurate angular position calculation.
The angular estimation by the Kalman filter consider the model described in (9) and (13), where
𝛥𝜃(𝑡) = 𝜃
̂𝑎(𝑡) − 𝜃
̂𝑔(𝑡), is the difference between the angle calculated from the accelerometer and the angle
calculated from the gyroscope. The noise produced in the acceleration measurement is modeled by 𝜂𝑎(𝑡).
The 𝑄𝑛 and 𝑅𝑛 are weighting matrices that represent the noise covariance in the states and output signal of
the model according to:
𝑄𝑛 = [
𝜎𝑔
2
0
0 𝜎𝑏𝑔
2] , 𝑅𝑛 = 𝜎𝑎
2
(20)

7.  ISSN: 1693-6930
TELKOMNIKA Telecommun Comput El Control, Vol. 21, No. 2, April 2023: 390-399
396
The parameters of the model in (9), (13), and (20) are shown in Table 1. The angular position
estimation for each sensor is performed according to Kalman’s classic filter equation:
𝑥
̂(𝑘 + 1) = 𝐹𝑥
̂(𝑘) + 𝐿(∆𝜃(𝑘) − 𝐶𝑥
̂(𝑘)) (21)
Where 𝑥
̂(𝑘) = [𝛥𝜃
̂(𝑘) 𝛥𝑏
̂𝑔(𝑘)]
′
are the states estimated by the Kalman filter, 𝐹 is the equivalent of matrix 𝐴
for the discrete-time with the sample-time 𝑇𝑠, 𝑘 is the discrete-time iteration and 𝐿 is de Kalman vector.
Finally, the estimate of the angular position using sensory fusion is calculated in (22) and Table 2.
shows the Kalman vector for each link.
𝜃
̂(𝑘) = 𝜃𝑥𝑔(𝑘) + ∆𝜃
̂(𝑘|𝑘) (22)
Figure 6 shows the result of the proof of concept developed to validate the sensor network. Each figure
at the top shows signals for ∆𝜃 (blue) and ∆𝜃
̂ (red) used to estimate the angle of the respective link. At the
bottom, each figure shows the estimation by the Kalman filter (KF in blue color) compared to a standard
complementary filter (CF in red color). Angular position estimations for foot, shin, thigh, and hip are shown in
Figure 6(a) to Figure 6(d) respectively. In this experiment, results indicate that for ten-meter walking test is clear
that the two estimates are similar. However, the estimates obtained with KF result in lower amplitude peaks.
Figure 5. Estimation of the foot angular position by gyroscope and accelerometer signals
Figure 6(a) shows the result of estimating the angle of the foot. The sensor was placed on the user’s
shoe so that the starting angle is -18.8° relative to the 𝑥 coordinate axis. The maximum and minimum angular
position values obtained with the Kalman filter were 8.62° and -52.36°, respectively. The maximum and
minimum angular position values obtained with the complementary filter were 16.76° and -61.48°,
respectively. Figure 6(b) shows the result of estimating the angle of the shin. The starting angle is 80.5°
relative to the 𝑥 coordinate axis. The maximum and minimum angular position values obtained with the
Kalman filter were 103.96° and 25.8°, respectively.
The maximum and minimum angular position values obtained with the complementary filter were
16.76° and -61.48°, respectively. Figure 6(c) shows the result of estimating the angle of the thigh. The starting
angle is 67.5° relative to the 𝑥 coordinate axis. The maximum and minimum angular position values obtained
with the Kalman filter were 80.4° and 45.59°, respectively. The maximum and minimum angular position
values obtained with the complementary filter were 82.66° and 38.51°, respectively. Figure 6(d) shows the
result of estimating the angle of the hip. The starting angle is 59.22° relative to the 𝑥 coordinate axis.
The maximum and minimum angular position values obtained with the Kalman filter were 65.73° and
49.14°, respectively. The maximum and minimum angular position values obtained with the complementary
filter were 67.51° and 46.82°, respectively.
Table 1. Parameters for Kalman filter calculation for each inertial sensor
Kalman filter 𝜎𝑔 𝜎𝑏𝑔
𝜎𝑎 𝜏𝑔
Hip 0.0103 30 46.56 839.7
Thigh 0.0017 50 329.08 2138.5
Shin 0.0008 96.4 100.94 1465.3
Foot 291.2 18.594 291.29 3567.5

8. TELKOMNIKA Telecommun Comput El Control 
Wearable sensor network for lower limb angle estimation in robotics applications (Andres L. Jutinico)
397
(a) (b)
(c) (d)
Figure 6. Results of estimating angular position for: (a) foot, (b) shin, (c) thigh, and (d) hip
Table 2. Kalman vector for each link
Kalman filter Kalman’s constant: 𝐿
Hip [
0.0137
0.0006
]
Thigh [
0.0055
0.0015
]
Shin [
0.0138
0.0095
]
Foot [
0.0113
0.0064
]
4. CONCLUSION
This article shows a sensor network for angle estimation of lower limbs. We highlight that this network
configuration is proper for robotic rehabilitation environments since devices are wearables and wireless. This
network has an ESP8266 microcontroller for processing and communication in each node. The communication
between nodes and the server is carried out by the standard 802.11 b/g/n. We use the client-server architecture
with UDP protocol which allows a sample time of 10 ms. The Wi-Fi use allows an intuitive connection and
stable communication between nodes and the server. We performed a proof of concept to validate the
implemented sensor network and used the Kalman filter to estimate the angular position of the foot, shin,
thigh, and hip. The Kalman filter allowed sensor fusion from angular velocity and acceleration measurements
by the MPU6050 in each node. In this sense, the Kalman filter estimates the angle error due to the numerical
integration of angular velocity given by the gyroscope. Also, it diminishes the noise from the angular position
coming from the accelerometer. Therefore, we compensate the actual measurement to obtain an accuracy
angle. Results show that this sensor network is suitable for real-time robotic applications. In future works, we
hope to use this sensor network to combine kinematic and electromyographic signals in applications for
rehabilitation.

9.  ISSN: 1693-6930
TELKOMNIKA Telecommun Comput El Control, Vol. 21, No. 2, April 2023: 390-399
398
ACKNOWLEDGEMENTS
The authors are grateful to the Universidad Distrital Francisco José de Caldas. This work is
supported by Universidad Distrital Francisco José de Caldas under project 2-5-617-20.
REFERENCES
[1] V. L. Feigin et al., “Global, regional, and national burden of stroke and its risk factors, 1990–2019: a systematic analysis for the
Global Burden of Disease Study 2019,” The Lancet Neurology, vol. 20, no. 10, pp. 795–820, 2021, doi: 10.1016/S1474-
4422(21)00252-0.
[2] V. L. Feigin et al., “World Stroke Organization (WSO): Global Stroke Fact Sheet 2022,” International Journal of Stroke, vol. 17,
no. 1, 2022, doi: 10.1177/17474930211065917.
[3] D. S. -Méndez, M. A. -Montiel, and E. L. -González, “Design of an Exoskeleton Prototype for Shoulder Rehabilitation,” Revista
Mexicana de Ingenieria Biomedica, vol. 38, no. 1, pp. 330–342, 2017. [Online]. Available:
https://www.scielo.org.mx/pdf/rmib/v38n1/2395-9126-rmib-38-01-330.pdf
[4] F. M. Escalante, A. L. Jutinico, J. C. Jaimes, M. H. Terra, and A. A. G. Siqueira, “Markovian Robust Filtering and Control
Applied to Rehabilitation Robotics,” IEEE/ASME Transactions on Mechatronics, vol. 26, no. 1, pp. 491-502, 2021,
doi: 10.1109/TMECH.2020.3034245.
[5] K. Nizamis, A. Athanasiou, S. Almpani, C. Dimitrousis, and A. Astaras, “Converging Robotic Technologies in Targeted Neural
Rehabilitation: A Review of Emerging Solutions and Challenges,” Sensors, vol. 21, no. 6, 2021, doi: 10.3390/s21062084.
[6] C. A. V. -Romero, J. E. B. -Jaimes, and D. C. P. -González, “Configuration Parameters in Module XBEE-PRO® ZB S2B for
Measuring Environmental Variables,” Tecnura, vol. 19, no. 45, pp. 141–157, 2015, doi: 10.14483/udistrital.jour.tecnura.2015.3.a011.
[7] C. Cifuentes, A. Braidot, L. Rodríguez, M. Frisoli, A. Santiago, and A. Frizera, “Development of a wearable ZigBee sensor
system for upper limb rehabilitation robotics,” in 2012 4th IEEE RAS & EMBS International Conference on Biomedical Robotics
and Biomechatronics (BioRob), 2012, pp. 1989-1994, doi: 10.1109/BioRob.2012.6290926.
[8] Y. Liao, M. S. Leeson, Q. Cai, Q. Ai, and Q. Liu, “Mutual-information-based incremental relaying communications for wireless
biomedical implant systems,” Sensors, vol. 18, no. 2, 2018, doi: 10.3390/s18020515.
[9] D. Goyal and Sonal, “Power management in Wireless Sensor Network,” in 2016 3rd International Conference on Computing for
Sustainable Global Development (INDIACom), 2016, pp. 598-601. [Online]. Available: https://ieeexplore.ieee.org/document/7724333
[10] K. Sohraby, D. Minoli, and T. Znati, Wireless sensor networks: Technology, Protocols, and Applications, Hoboken, NJ, USA:
Wiley Telecom, John Wiley & Sons, Inc., 2007. [Online]. Available: https://www.wiley.com/en-
us/Wireless+Sensor+Networks%3A+Technology%2C+Protocols%2C+and+Applications-p-9780470112755
[11] A. U. Alahakone and S. M. N. A. Senanayake, “A combination of inertial sensors and vibrotactile feedback for balance
improvements in therapeutic applications,” in Proc. 2009 Innovative Technologies in Intelligent Systems and Industrial
Applications, CITISIA 2009, 2009, pp. 5–10, doi: 10.1109/CITISIA.2009.5224252.
[12] A. Kos and A. Umek, “Reliable Communication Protocol for Coach Based Augmented Biofeedback Applications in Swimming,”
Procedia Computer Science, vol. 174, pp. 351–357, 2020, doi: 10.1016/j.procs.2020.06.098.
[13] A. Johansson, W. Shen, and Y. Xu, “An ANT based wireless body sensor biofeedback network for medical E-health care,” in
Proc, 7th International Conference on Wireless Communications, Networking and Mobile Computing, 2011, pp. 1-5,
doi: 10.1109/wicom.2011.6040656.
[14] M. Benocci, E. Farella, L. Benini and L. Vanzago, “Optimizing ZigBee for data streaming in body-area bio-feedback applications,”
2009 3rd International Workshop on Advances in sensors and Interfaces, 2009, pp. 150-155, doi: 10.1109/IWASI.2009.5184786.
[15] L. Maggi, S. Parini, L. Piccini, G. Panfili, and G. Andreoni, “A four command BCI system based on the SSVEP protocol,” in
2006 International Conference of the IEEE Engineering in Medicine and Biology Society, 2006, pp. 1264-1267, doi:
10.1109/IEMBS.2006.260353.
[16] L. G. -Villanueva, S. Cagnoni, and L. Ascari, “Design of a wearable sensing system for human motion monitoring in physical
rehabilitation,” Sensors, vol. 13, no. 6, pp. 7735–7755, 2013, doi: 10.3390/s130607735.
[17] K. M. Bell et al., “Verification of a portable motion tracking system for remote management of physical rehabilitation of the
knee,” Sensors, vol. 19, no. 5, 2019, doi: 10.3390/s19051021.
[18] M. T. -Quezada, R. S. -Zamora, L. B. -Vázquez, D. D. -Rodríguez, and A. L. -Delis, “Robotic training system for upper limb
rehabilitation,” Ingenieria y Universidad, vol. 18, no. 2, pp. 235–252, 2014, doi: 10.11144/Javeriana.IYU18-2.rtsu.
[19] A. R. D. -Elli and P. G. Adamczyk, “Design and Validation of a Lower-Limb Haptic Rehabilitation Robot,” IEEE Transactions
on Neural Systems and Rehabilitation Engineering, vol. 28, no. 7, pp. 1584–1594, 2020, doi: 10.1109/TNSRE.2020.3000735.
[20] G. G. Peña, L. J. Consoni, W. M. dos Santos, and A. A. G. Siqueira, “Feasibility of an optimal EMG-driven adaptive impedance control
applied to an active knee orthosis,” Robotics and Autonomous Systems, vol. 112, pp. 98–108, 2019, doi: 10.1016/j.robot.2018.11.011.
[21] F. M. Escalante, J. C. P. -Ibarra, J. C. Jaimes, A. A. G. Siqueira, and M. H. Terra, “Robust Markovian Impedance Control applied
to a modular knee-exoskeleton applied to a modular knee-exoskeleton,” IFAC-PapersOnLine, vol. 53, no. 2, pp. 10141–10147,
2020, doi: 10.1016/j.ifacol.2020.12.2740.
[22] A. L. Jutinico, O. F. -Cediel, and A. A. G. Siqueira, “Complementary stability of Markovian systems: Series elastic actuators and
human-robot interaction,” IFAC-PapersOnLine, vol. 54, no. 4, pp. 112–117, 2021, doi: 10.1016/j.ifacol.2021.10.019.
[23] R. L. -Gutiérrez, H. A. -Sierra, S. Salazar, and R. Lozano, “Adaptive control in passive rehabilitation routines using ELLTIO,”
Revista Mexicana de Ingenieria Biomedica, vol. 38, no. 2, pp. 458–478, 2017, doi: 10.17488/RMIB.38.2.3.
[24] M. Sartori, M. Maculan, C. Pizzolato, M. Reggiani, and D. Farina, “Modeling and simulating the neuromuscular mechanisms
regulating ankle and knee joint stiffness during human locomotion,” Journal of Neurophysiology, vol. 114, no. 4, pp. 2509–2527,
2015, doi: 10.1152/jn.00989.2014.
[25] S. L. Nogueira, A. A. G. Siqueira, R. S. Inoue, and M. H. Terra, “Markov Jump Linear Systems-based position estimation for
lower limb exoskeletons,” Sensors, vol. 14, no. 1, pp. 1835–1849, 2014, doi: 10.3390/s140101835.
[26] S. L. Nogueira et al., “Global Kalman filter approaches to estimate absolute angles of lower limb segments,” BioMedical
Engineering Online, vol. 16, no. 58, 2017, doi: 10.1186/s12938-017-0346-7.

10. TELKOMNIKA Telecommun Comput El Control 
Wearable sensor network for lower limb angle estimation in robotics applications (Andres L. Jutinico)
399
BIOGRAPHIES OF AUTHORS
Andres L. Jutinico received the bachelor’s in electronic engineering from The
Universidad Distrital Francisco José de Caldas, Bogota, Colombia, in 2005, the M.Sc. degree
in industrial automation from the National University of Colombia, in 2012, and the Ph.D.
degree in mechanical engineering from the University of São Paulo, Brazil, in 2019. His
current research interests include control systems, assistive and rehabilitation robotics, and
sensor fusion for assistive technology. He can be contacted at email:
aljutinicoa@udistrital.edu.co.
Gustavo A. R. Rodríguez received the bachelor’s in electronic engineering from
The Universidad Distrital Francisco José de Caldas, Bogota, Colombia, in 2015, the M.Sc.
information and communication sciences with an emphasis on teleinformatics in 2022. His
current research interests include control systems, assistive and rehabilitation robotics, and
sensor fusion for assistive technology. He can be contacted at email:
garodriguezr@correo.udistrital.edu.co.
Julian Rolando Camargo Lopez is an Associate Professor of Faculty of
Engineering at Universidad Distrital Francisco José de Caldas in Bogotá, Colombia. Electronic
Engineer from the Universidad Distrital Francisco José de Caldas, Specialist in Design and
Construction of Telematic Solutions from the Autonoma University of Colombia, master’s in
information and communications sciences from Universidad Distrital Francisco José de
Caldas. He can be contacted at email: jcamargo@udistrital.edu.co.